A growing number of applications and devices make use of speech recognition functions to enhance the usability of the applications. However, speech recognition applications and resources can be expensive to create, and may be too resource-intensive to fit on smaller portable devices. Network hosted speech recognition services may have application program interfaces (APIs) that may require a high level of skill to integrate with other applications. Some speech recognition services may not work across engine implementations, or device boundaries, e.g. desktop, server, or embedded.
It is with respect to these and other considerations that the present improvements have been needed.